Determining Temperature Profiles in the Extruder during the Material Extrusion of an Amorphous Polymer Using Two Coupled Physics-Informed Neural Networks

There is no exact analytical solution for the heat-transfer problem in the extruder of material extrusion additive manufacturing, except for approximate solutions to simplified cases. Physics-informed neural networks (PINNs) are machine learning models that use artificial neural networks to numerically solve problems based on underlying physical laws. In this work, two coupled PINNs are developed to solve the heat-transfer problem without simplifying the governing equations or interface conditions. The goal is to accurately determine the temperature profiles inside the extruder, focusing specifically on the extrusion of amorphous polymers. To validate the approach, the two PINNs are first applied separately to solve simpler heat-transfer problems. They are then combined to solve the full heat-transfer problem. The PINN solution is compared with the approximate solutions to simplified cases for extruding acrylonitrile butadiene styrene (ABS). The Python implementation runs in under 10 min on an online platform, demonstrating practical usability. The developed PINN model provides a simple and efficient method for determining temperature profiles in the extruder during the extrusion of various amorphous materials, delivering critical information for optimizing the MatEx process.